In August 2006 AstraZeneca completed the ARISE trial, which aimed to determine whether AGI-1067 was effective in reducing atherosclerosis in patients with acute coronary artery disease . A double-blind, placebo controlled, Phase 3 trial, the primary efficacy endpoint was a composite endpoint which included major adverse cardiovascular events (MACE) like cardiovascular death, resuscitated cardiac arrest, non-fatal myocardial infarction and non-fatal stroke. However, it also included two less serious but more frequently observed events, namely hospitalization due to coronary revascularization and hospitalization due to unstable angina with evidence of ischemia .
These last two endpoints, though less serious than the other four, had far more frequent occurrence.Unfortunately, the addition of these more frequently occurring events to the primary composite endpoint eventually led to the trial’s failure.
As it happens, AGI-1067 diminishes certain MACE events but results in greater frequency of hospitalizations. The trial found 530 adverse events in the active arm and 529 in the placebo.By contrast, the composite secondary endpoint, which only consisted of the four more serious events, clearly demonstrated the beneficial effects of the new therapy with a 19% risk reduction in these areas .
A frequently observed event has obvious benefits for a clinical trial. It increases the number of events observed, thereby increasing the study’s power and also accelerating study completion. In the case of the ARISE trial, despite enrolling 6144 patients, the trial completed in twenty-five months .
However, increasing the types of events observed does not guarantee trial success. In cases like the ARISE trial, this strategy could cause serious harm in trying to demonstrate efficacy.
In the past few years, statisticians have begun to provide a range of strategic options for constructing more successful composite endpoints. Stay tuned to learn more…
Related Items of Interest
 Gómez, Guadalupe, and Stephen W. Lagakos. "Statistical considerations when using a composite endpoint for comparing treatment groups." Statistics in medicine 32.5 (2013): 719-738.